[1M3-GS-13] AI application: Social data and prediction

[1M3-GS-13-01] Landmark Detection on Ultrasound Image Using Deep Learning

〇Yurie Kanauchi¹, Masahiro Hashimoto², Takumi Seto¹, Haque Hasnine³, Masahiro Jinzaki², Yasubumi Sakakibara¹ (1. Faculty of Science and Technology, Keio University, 2. Keio University School of Medicine, 3. GE Healthcare Japan)

Keywords:ultrasound imaging, landmark detection, deep learning

In medical imaging, landmarks have significant clinical and scientific importance. Clinical measurements, derived from the landmarks, are used for diagnosis. However, placing landmarks manually is a tedious task. In ultrasound imaging, measurement marker is generally used to mark a pair of landmarks by the operator to measure the geometry of the region of interest. Manual placement of the measurement markers is subject to inter-observer variability and there is a need to make the process automatic. The aim of this study is to develop a new landmark detection method that the domain is not limited using machine learning technique. The neural network model was designed and trained using about 50,000 images of various domains so that if one of the measurement markers is given as input, the second measurement marker position will be predicted. We evaluated the prediction accuracy and to confirm its usefulness.